The framework categorizes all government AI initiatives into four quadrants: internal agents, internal RAG systems, external agents, and external RAG systems. This structured approach helps spot duplication, clean up investment decisions, and keep the system coherent.